Documente Academic
Documente Profesional
Documente Cultură
of tasks'. This rather general definition of the term does convey two different meanings that
schedule. In this sense, much of what we learn about scheduling can apply to other kinds of
techniques, and logical conclusions that provide insight into the scheduling function.
The vital elements in scheduling models are resources and tasks. The set of tasks
available for scheduling does not change over time, the system is called static', in contrast
to cases in which new tasks arise over time, where the system is called dynamic^.
and it is a decision making function. The practical problem of allocating resources over
however, scheduling does not become a concern until some fundamental planning problems
are resolved, and it must be recognized that scheduling decisions are of secondary
importance to a broader set of managerial decisions. The scheduling process most often
arises in a situation where resource availability fixed by the long- term commitments of a
prior-planning horizon.
whenever there is a choice to the order in which a group of tasks can be performed. The
shop supervisor or scheduler can deal with sequencing problems in a variety of ways. The
simplest approach is to ignore the problem and accomplish the tasks in any random order.
"rules of thumb". In certain cases, scientifically derived scheduling procedures can be used
, This section addresses some of the objectives of job shop scheduling environment.
For a finite set of tasks, the utilization of the resources is inversely proportional to
the time required to accomplish all the tasks. This time is known as the makespan or
maximum flow time of the schedule in a static scheduling system. In finite scheduling
makespan".
27
2.2.2 Minimization of the Worli-in-process Inventory :
The reduction of the average number of tasks waiting in a queue while the resources
is busy with other tasks. If the jobs spend less time in the system, the inventory is reduced.
In scheduling terms, this leads to minimizing the average flow time (mean flow time) or the
average of the times the jobs spend in the system. The makespan of a scheduling is
constant, the sequence that reduces mean flow time also reduces mean in-process
inventory^.
In many situations, some or all of the tasks have due dates and a penalty is incurred
if a task is finished after that due date. The possible objectives relating to tardiness are the
minimization of maximum tardiness, the minimization of the number of tasks that are tardy,
or the minimization of mean tardiness . Sometimes, completion of the jobs ahead of the due
dates is also undesirable. For such scheduling problems, the objective is to minimize a
penalty function of earliness and tardiness. Here, it can minimize either the number of
early/tardy jobs or minimize the maximum earliness/tardiness. if all the jobs have a
common due date, this is usually achieved by minimizing the squared deviation of the job
completion times about the due date. If the analyst can fix the due date, the problem
to maximize customer safisfaction, maximize shop utilization and minimize operating costs.
28
> Sequencing
> Routing
are: Job shop, open job shop, batch shop, flow shop, batch/flow shop, manufacturing cell,
The setting in which shop scheduling is done is described below: Each job
represents one or more tasks that must be performed in a prescribed order and can be
assigned to one or more machines, which have the same or different processing routes.
Scheduling consists of two stages, viz. loading and sequencing. The first stage involves
quality, setup costs, preventive maintenance, and operation availability may dictate the job
assignment. All else being equal, a good rule would be to assign a job to the machine with
the least workload. The second stage of job scheduling is to sequence the tasks on the
The most elementary' scheduling problem occurs whenever a set of tasks is waiting
to be accomplished and only one processor is available. The processing times and due date
of each job are known and are independent of the sequence in which the tasks are run. The
scheduling problem in this situation is one of deciding the order in which tasks are to be
accomplished. The choice of sequence will fix the completion time of each task. The
29
makespan, which is the time needed to complete the whole set of tasks, is constant and it is
independent of the sequence. If it assume that all tasks are available when the schedule is
started, the flow time of each task equals its completion time. Although the makespan is a
constant for all the sequences, the mean flow time, mean lateness, and mean tardiness they
can be minimized by selecting a good sequence. It is well known that the Shortest
In the case of m parallel processors, each job visits only one of the processors. In the
case of m processors in series, each task must visit each processor in the same order. In the
case of m identical processors, the problem is to select both the processor to be used and the
sequence for the tasks on each processor. If the objective is to minimize mean flow time, a
simple variation of the SPT scheduling rule can be used. Scheduling situations with two or
more identical processors in parallel can be handled by first ordering all tasks using some
appropriate single-processor rule and then allocating the tasks to the processor with the least
scheduled time.
The classical Job Shop Scheduling (JSS) problem can be stated as follows. There
are m different machines and n different jobs to be scheduled. Each job is composed of a
set of operations and order on each machine is prespecified. The required machine and the
fixed processing time characterize each operation. In the job shop case it is more
appropriate to describe an operation with a triplet (i, j , k) in order to denote the fact that
30
operation j of job i requires machine k. Workflow at a typical machine in a job shop is
Figure: 2.1
NEW JOBS
^r
MACHINE K ^
^ COMPLETED JOBS
^
OPERATION NUMBER
(i , j ^, k )
/ ^
In principle, there are infinite numbers of feasible schedules for any job shop
problem, because an arbitrary amount of idle time can be inserted at any machine between
adjacent pair of operations. It should be clear, however, that once the operation sequence is
specified for each machine, its kind o^idle time couldn't be helpful for any regular measure
31
of performance. Rather, it is desirable that the operations can be processed as compactly as
possible. Superfluous idle time exists in a schedule if some operations can be started earlier
in time without altering the operation sequences on any machine. Adjusting the start time
of some operation in this way is equivalent to moving an operation block to the left on the
Gantt chart'^ while preserving the operation sequences. This type of adjustment is called a
local left-shift.
Given an operation sequence for each machine, there is only one schedule in which
no local shift can be made. The set of all schedules in which no local shift can be made is
called the set of semi active schedules and is equivalent to the set of all schedules that
contains none of the superfluous idle time. This set dominates the set of all schedules,
which means that it is sufficient to consider only semi active schedules to optimize any
The number of semi active schedules is at least finite, although it may well be quite
large. The exact number is usually difficult to determine. For the classical job shop
scheduling problem, in which each job has exactly one operation on each machine, each
machine must perform n operations. The number of possible sequences is therefore n! for
each machine. If the sequences on each machine were entirely independent, there would be
(n!)'" semi active schedules. However, the effect of the precedence structure and machine
routing for each job is usually to render some of these combinations of sequences are
infeasible.
Just as the set of semi active schedules dominates the set of all schedules. In
schedules'". Evidently the number of active schedules is a function of both the routing and
32
the processing times in a given problem, whereas the number of semi active schedules is a
machine sequences, as discussed above, many semi active schedules can often be
compacted into the same active schedule through a series of left-shifts. Several different
active schedules can be constructed by a series of left-shifts starting with a given semi
active schedule. The roles of semi active and active schedules are shown in the Figure. 2.2
Figure: 2.2
The large rectangle in the Figure. 2.2 represents the roll of semi active and an
Active schedule represents the set of all the schedules. The interior of the region labeled S-
A represents the finite set of semi active schedules. Wholly contained within that set is the
set of active schedules, represented by the region labeled A. The asterisk represents some
optimal schedules, placed to indicate that at least one optimum must lie in the active subset.
The number if active schedules still tends to be large, and it is some times
convenient to focus on an even smaller subset called the non-delay schedules. In a non-
delay schedule no machine is kept idle at a time when it could begin processing some
operation.
All non-delay schedules are active schedules since no left shifting would be
possible. One the other hand many active schedules may not be non-delay schedules. This
means that the number of non-delay schedules can be significantly less than the number of
active schedules. There is no guarantee that the non-delay subset will contain an optimum.
The Figure.2.3 depicts a situation in which at least one optimal schedule, while the
34
Figure:2.3
Figure:2.4
35
The active schedules are generally the smallest dominated set in the job shop
problem . The non-delay schedules are smaller in number but are not dominant. The best
non-delay schedules can usually be expected to provide a very good solution, if not an
optimum. In this sense, the role of non-delay schedules is similar to that of the permutation
schedules in large flow shop problems: although the set is not always dominant, it can
One basic distinction in scheduling research refers to the nature of the job arrivals in
the shop. In a static model jobs arrive simultaneously and are available for to be scheduled
at the same instant. Accordingly, their ready or release times are zero. i.e. the total set of
A dynamic model allows for a continuous stream of arriving orders in time that are
intermittently released to the shop and are included in the current scheduling procedure.
Reasonably, the distinction between simultaneous and intermittent job arrivals involves the
one between known and fixed job data on one hand and stochastic data, in particular job
dynamic/stochastic one. Here, the emphasis is on the both cases. Within the subset of
ignoring the analytical procedures by means of queuing theory systems. The vast majority
36
distribution''* of job arrivals and, processing time is considered as random variable,
scheduling approach is due to the fact that it allows for an up-to-date decision with respect
to meanwhile entering (possibly rush) jobs, by loading a machine at the latest possible
moment, namely as that machine gets idle. Contrarily, a static model postpones the urgent
disadvantage that each sequencing decisions based on the constrained information horizon
given by the set of currently schedulable jobs, which prohibits the definition of an overall
optimum: due to the real-time capability of a given sequencing decision, only the selection
of the first job of the computed job sequence on that machine is actually performed, while
the remainder of the scheduled jobs is rescheduled and possibly revised on the occasion of
the subsequent loading moment. Thus, rather than to determine a global optimal sequencing
policy, a dynamic job shop scheduling simulafion'"'' at best is able to provide a heuristic
optimum among alternative sequencing strategies by which a given job file is scheduled
through the shop in successive simulation runs such a policy that defines a specific
sequencing decision each time a machine gets idle, is called a priority rule.
A priority rule allows an idle machine to select its next operafion fi-om among those
machine; but the availability-property may also be extended to jobs being currently in the
queue or on the machine of other workstations before proceeding to the queue in question.
37
^ Mahendra, S. Bakshi., Sant Ram Arora. 'T/ze sequencing problem" Manaeement
Science 16 No. 4 (1969): pp 246-263
^ Fam, C.K., Muhlemann, A.P. " The dynamic aspects of a production scheduling
problem " International Journal of Production Research 17 No. 1 (1979): pp 15-21.
'* Kenneth R. Baker., Introduction to sequencing and scheduling, John Wiley & Sons, Inc.:
1974.
"''Hemant Kumar, N.S., Srinivasan, G., "A genetic algorithm for Job shop scheduling-
A case study " Computers in Industry 31 (1996): pp 155-160.
^ Michael pinedo, Scheduling theory, algorithms and systems, Prentice Hall Englewood
Cliffs, New Jersey: 1995.
^ Douglas A. Elvers., ""The sensitivity of the relative effectiveness of job shop dispatching
rules with respect to various arrival distributions" AllE Transactions 6 No. 1 (1974):
pp 41-49.
Muhlemann, A.P., Lockett. A.G., and Fam. C.K.. "Job shop scheduling heuristics and
frequency of scheduling "International Journal of Production Research 20 No.2 (1982):
pp 227-241.
Haupt, R., " A sun'ev ofprioritv rule based scheduling " Operations Research Spektrum
11 (1989): pp 3-16. '
Kannan, V.R., Ghosh, S., ""An evaluation of the interaction between dispatching rules and
truncation procedures in job shop scheduling" International Journal of Production
Research 31 No.7(1993): pp 1637-1654.
Simon French, Sequencing and scheduling- An introduction to the mathematics of the job
shop, Ellis Horwood Ltd., England: 1987.
Eilon, S., Cotterill. D.J., ""A modified SI rule in job shop scheduling" International
Journal of Production Research 7 No. 2(1968): pp 135-145.
Sun, D.. Batta, R.. " Scheduling larger job shops: A decomposition approach"
International Journal of Production Research 34 No.7 (1996): pp 2019-2033.
'" Michael, H. Bulkin.. John, L. Colley., and Harry. W. Steinhoff.JR., " Load forecasting,
priority sequencing and simulation in a job shop control svstem " Management Science 13
No.2 (1996): pp B-28-B51.
" Doctor, S.R., Cavalier. T.M., Egbelu, P.J., ""Scheduling for machining and assembly in a
job shop environment" International Journal of Production Research 31 No. 6 (1993):
pp 1275-1297.
38